[PATCH] Propagate -fno-unwind-tables to EHABI

[Renato Golin]

  > BTW, I found that I was wrong in the previous comment. If we wish to keep the test cases unchanged, then we should emit unwinding information for the functions with neither uwtable nor nounwind attributes.

  The whole point of this patch is to change that, so that -fno-unwind-tables can remove them. I cannot understand why would I want to emit tables when no unwind information is present. Or, better phrased, why would you need uwtable if on every case the tables are emitted.  For me, that behaviour was just wrong, and was created when I enabled EHABI by default, so that's not what we wanted (or had) in the first place.

  This may be the case for x86_64, but it's not for ARM. We can't model all EH emissions based on x86_64's ABI.

http://llvm-reviews.chandlerc.com/D3079